Review of EGUSPHERE-2025-5149: Aerosol-deep convection interaction based on joint cell-thermal tracking in Large Eddy Simulations during the TRACER campaign

General Comments:

In this study, the authors run two simulations at high-resolution (dx=dy=200m) of two cases during the TRACER field campaign. They initialize these simulations with clean and polluted aerosol conditions, based on observations from TRACER. The thermals within these simulations are tracked to assess the impacts of aerosol impacts on deep convective cloud processes on a thermal-centric level. Generally, their figures and discussion were well-presented and easy-to-follow. Their thermal-centric view is very compelling and allows for a more detailed assessment of aerosol impacts on finer scales than prior studies, which adds new insights to specific processes, particularly with respect to the locations of hydrometeors with respect to the thermal centers. This focus on hydrometeors is particularly helpful given that this study also focuses on lightning. The authors’ conclusions generally support results shown in the literature with different modeling frameworks, which is encouraging. While I appreciate the authors’ additional focus on environmental feedbacks, which are important, most of my more significant questions and concerns are related to their analyses and experimental set-up with regards to the environmental feedbacks. Additionally, some other clarifications, including details about their thermal sample composites and the aerosol evolution, would strengthen the manuscript.

Specific Comments:

• L100: I appreciate that the authors constrain their analyses with realistic aerosol conditions from observations.
• L105-107: How well do the log-normal distributions with the parameters listed in Table 1 fit the TRACER observations? A quantification of this fit or an image that shows the fit would be helpful for the reader.
• L126: Can the authors provide an estimate of the vertical grid spacing, as some readers may not be able to easily estimate how 91 model levels translate to vertical distance?
• L149-150: How was this vertical distribution of aerosol decided? Was it also based on observations from TRACER?
• L168: How are the cell merger and splits accounted for. Are they removed? The authors state that the tracks are “refined” based on the evaluation of mergers and splits, but this is vague. Can the authors be more explicit here?
• L170: What is a neighboring feature? This was not defined, so it is hard to understand this limitation on their tracks.
• L206: It is great that the authors consider more than one case to create more robust results.
• L225: The authors use comparative language (e.g., relatively), but it is unclear what they are comparing against.
• L238-240: The authors track thermals within the tracked cells for each of the two cases. For the clean simulations, ~33% of thermals come from 08/04 case and 67% of thermals come from the 08/07 case. However, for the polluted simulations, only 15% of the thermals come from the 08/04 case, while 85% are from the 08/07 case. As such, differences in polluted and clean simulations may be impacted by the different data weights from these cases. Have the authors checked this sampling impact on their results?
• Figure 4: Before diving into the results of clean and polluted thermals, it would be helpful to understand how the aerosol concentration profiles evolved during the simulation. It seems like much of the analysis is focused on the last 6-12 hours of their simulation, after 12-18 hours of model integration. As such, during these last 6-12 hours, do the differences in clean versus polluted match those shown in Table 2 from the initialization, and if not, what are the aerosol differences at these later times? This would be helpful in terms of understanding the actual aerosol perturbation being analyzed in this study and help with the sensitivity experiments described later.
• Figure 5: There seems to be very few thermals that initiate below 2 km (panel d). I understand their reflectivity threshold on cells may miss the early stages of the convective lifecycle; however, for more mature cells, are thermals not originating in the boundary layer? Can the authors explain the lack of thermals below 2 km AGL?
• L252 / L312: The thermal analyses weight their composite by mass flux, which emphasizes the strongest thermals. I am curious how sensitive their findings are to this weighting and whether an equal weighting (e.g., regular mean and median) produces similar results or provides different insights on aerosol impacts on thermals.
• Figure 8: I really like the analyses in and evolution of Figures 4, 7 and 8.
• Figure 11: It seems like having a time series of a reflectivity or rain rate or updraft contour in panels a, b, e, f, such that the time evolution of updraft or precipitation would be more comparable to the time evolution of the Meso-LLC in panels c, d, g, h.
• L373-381: The Aerosol-Cloud-Precipitation-Climate Model Intercomparison Project, which focused on the Houston region with similar sea breeze convection, also found a very similar trend in the environmental response of temperature and moisture at the same vertical levels described in this section. As such, I think it would be helpful to cite that work (Marinescu et al., 2021) here, which will help demonstrate the robustness of this result.
• L391: 0 and 5 km AGL seems like a very deep layer in terms of understanding low-level wind convergence. Why do the authors use 5km AGL as an upper limit? Could other thermal circulations above the bottom (inflow) of the thermal be included in these results due to using such a high limit?
• L394: Are the mesoscale low-level convergence features discussed here referring to cold pools, a sea breeze front or some other low-level convergence boundary? Some clarification here would be helpful.
• L394: Additionally, depending on what the feature is causing the mesoscale low-level conference in this case (e.g., sea-breezes, cold pools), tying in some of the seminal work in terms of aerosol impacts on those specific mesoscale features may be helpful in terms of understanding the physical mechanism at play.
• L394-400: What is the physical mechanism by which aerosol particles impact the clustering differences of convective features within the different simulations? Can the authors speculate about this and/or relate this finding to prior research?
• L397-398: The authors state that under polluted conditions their cells / forcing are more tightly clustered and more consolidated based on a qualitative assessment of Figure 11, which is a snapshot at one time. There are many quantitative assessments of organization in the literature (e.g., I_org, L_org), and perhaps these could be used to provide a more quantitative assessment.
• L398: The authors state that the low level convergence is “markedly stronger”. Can the authors quantify this, as it is hard to determine this from Figure 11? Additionally, have the authors considered including convergence in their thermal-centric view plots (e.g., Figure 8), which I thought were very compelling analyses and would show the mesoscale dynamical feedbacks that more directly impact the thermals / updrafts?
• L402: In terms of understanding the mesoscale feedbacks and interactions, the sample size of what has been presented is effectively 2 simulation snapshots. While I recognize the computational limitations of running an ensemble at high resolution, I do think that it may be hard to draw conclusions on the impacts of mesoscale clustering based on this small sample size. This limitation should be clearly stated in this section and the conclusion.
• L407: The authors state that appreciable strengthening only occurs in the late stages, but what do they mean by appreciable? And what is being strengthened? Can they quantify and be more explicit?
• L409: Can they authors be a little clear on what hypothesis they are testing here. They mention a time-sequential impact on the mesoscale environment on the prior line, but what exactly does that mean?
• L409-418: The authors conduct sensitivity simulations where they re-initialize their aerosol concentrations at the beginning of a 2-hour period at the end of their simulations. How does their model deal with such a strong perturbation from one time step to the next? For example, a grid point with clean aerosol conditions is replaced by one with polluted aerosol conditions instantaneously, which could lead to unphysical responses, particularly right after the change. Did the authors consider allowing for some spin-up time to allow for unphysical conditions to be alleviated by the model?
• L409-418: How sensitive are the authors results to the specific timing of these sensitivity experiments? For example, if the aerosol reinitialization occurred at 18 UTC or 20 UTC, as opposed to 22 UTC, how would that impact the results? Similarly, how sensitive are the results to the length of time of analysis after the reinitialization (2 hours)? I assume these results would be sensitive to the lifecycle stage of the convection. I think it would be helpful if the authors can spend more time motivating these experiments since they seem somewhat ill-constrained.
• L428-432. The authors show that the PollRun experiment that was reinitialized with “Clean” aerosol conditions have similar ice and charge conditions as the other PollRun experiments in their domain and time averaged profiles. The authors conclude that this suggests that the mesoscale environment development in the PollRun experiment plays a dominant role in impacting these hydrometeor and electrification profiles since the simulation with reinitialized clean aerosol shows a similar response. However, since these are based on domain mean profiles and the lifetimes of cloud ice can extend for many hours, could this finding be due to the fact that these simulations were run with polluted conditions for ~10 hours before this with polluted conditions. I am having a hard time understanding how the authors can isolate that this is due to the environmental response as opposed to the integrated microphysical response.
• Figure 12: Given the focus on updraft thermals, it would be helpful to include vertical velocity as a panel in Figure 12.
• L433: The authors conduct the same reinitialization experiments for their second case but do not show the results. What is the reasoning for this? The addition of these results would be valuable in terms of improving the robustness of their sensitivity experiment analyses.
• L442: Are the differences/similarities in Figure 12 that are a result of the mesoscale environment due to changes in the temperature/moisture/stability or due to low-level convergence or both? The authors generally use the term “mesoscale environment,” and I think it would be helpful to make it clearer what they are specifically referring to.
• L450: The authors refer to these cases as “golden.” What do they mean by this?
• L483-491: As mentioned above, the ACPC MIP results from Marinescu et al., 2021 also found similar changes to the water vapor and temperature profiles in the environment (their Figure 5, panels i, j and resulting discussion) from a range of models for a similar case study (i.e., same region, type of convection). As such, it should be cited here to strengthen this finding.
• L492-494: Saleeby et al., 2025 focused on tracked updrafts and the microphysical process rates within aerosol perturbation experiments for similar convection in the same region with many different models in the ACPC MIP and could be used here to strengthen this finding about process rates.
• L509-510: Again, this finding about changes in the mesoscale environment using a regional LES domain within a 24-hour integration was also presented in the ACPC MIP studies (Marinescu et al., 2021).
• L511-517. The authors state that “at least, quantifications of changes in the mesoscale environment will be requirement for future MIP activities.” I apologize for bringing this up again, but it seems like the authors may not have been aware that this was done in the MIP studies that they reference here (Marinescu et al., 2021, Saleeby et al., 2025; and van den Heever et al., 2025). As noted in several comments in this review, Marinescu et al., 2021 does quantify changes to the environment within the ACPC MIP model simulations (their Figure 5) and discusses similar mesoscale, environmental feedbacks as discussed in this study to explain the aerosol-impacts on updrafts. Specifically, that study showed both warming and drying in the mesoscale environment boundary layer under more polluted conditions and stated explicitly in the conclusions section that “the consistent CCN-induced response [ in updrafts ] is likely related to an environmental feedback process where the High-CCN simulations have increased environmental instability as a result of warmer boundary layer temperatures and cooler cloud level temperatures.” That study also shows and has discussion about the drier boundary layer environments and moister cloud level environments in polluted conditions. Given that the authors research focuses on the same region and with similar convective features (e.g., scattered, sea breeze convection), the consistent findings between the authors’ work and the ACPC MIP strengthens the conclusion about aerosol impacts on the environment. Can the authors reword this section to accurately reflect this?

Technical Corrections:

• Figure 7 caption – supersaturation with respect to liquid or ice?
• Some of the figures’ text are small (e.g., Figure 9, 12), and I was wondering if the authors could make them larger.

Review for egusphere-2025-5149 titled “Aerosol-deep convection interactions based on joint cell-thermal tracking in Large Eddy Simulations during the TRACER campaign”

The authors use thermal tracking within large-eddy simulations of isolated deep convection for the DOE TRACER field campaign near Houston with an IOP in summer 2022. The authors employ a thermal tracking approach to “investigate the response of cold- and warm-phase microphysics and electrification processes within thermals to varying aerosol loading conditions during isolated deep convective events”. They simulate isolated convection for 2 dates in August using observed aerosol measurements from the campaign to address the aerosol impacts on cloud microphysics and convection aggregation.

This is a well-executed study that addresses key shortcomings in existing literature w.r.t. aerosol effects on deep convective cloud microphysics. The results are easy to follow, and the manuscript presents insights into aerosol-cloud microphysics interactions that can be built upon in future work. I recommend this manuscript be published after minor revisions are made to address the comments below.

I have a general question related to the applicability of the results to other cases from TRACER. Two dates are selected and there are some distinct patterns in convective activity between these dates already. There is also a lack of discussion of the observed convection on these dates (from NEXRAD or ARM radar scans) and the observed aerosol concentrations. Aspects related to convection aggregation can be placed in contrast with observations.

Given that thermals were tracked, the lifecycle evolution of the thermals is missing. See related comment below regarding the possibility of creating an additional figure similar to Figure 6 where multiple lines represent clean, polluted, and/or clean-polluted cases where each line represents the vertical profile of key thermal properties as a function of the normalized thermal lifetime. This is somewhat done with the full-domain time series and the composites for different altitude levels but those seem indirect ways of representing the thermal evolution when a better methodology is available to the authors given the tobac tracking.

Minor Comments:

Abstract: The abstract is hard to follow and lacks clarity from Line 10 onward. I suggest re-wording it based on the following comments:

• Line 11: “higher” and “faster” in this context must be clarified. Does “higher” refer to a thermal’s base altitude? Can you quantify the aerosol impacts as a relative change from cases with low aerosol concentrations?
• Line 12: This sentence needs to be rephrased.
• Line 13: Please specify the feedback.
• Line 14: “aerosol-reinitialization”?
• Line 15: Do you mean smaller spatial scale? Or shorter time scale?
• The abstract should mention that two case dates chosen for detailed analysis form the basis of these results.

Line 24: Suggest re-wording this sentence to be more specific.

Line 52: Suggest specifying for what is thermal size the most relevant determining factor

Line 100: Are these daily percentiles or for the entire period? Please specify either way. If latter, please provide the aerosol concentration values for each percentile.

Line 104: Does either the reference or the current study have a size range for each of the three aerosol modes? Can the ranges be provided here or in the table?

Line 129: Do the authors have any comment on why no PBL scheme was used and whether that might have an impact on the subsequent analyses?

Line 170: Can you provide the rationale for removing “anomalous” cells as defined here? The exclusion of aggregated cells based on “number of neighboring cells” warrants further clarity – was a distance threshold used? What was it? Did a cell have to have a certain number of cells within a threshold distance to be termed “aggregated”?

Line 206: Can you expand on what is meant by “careful investigation of these results”? Based on the current text, it seems the implication is that these dates saw a high (highest?) number of isolated cases compared to other days and the forecasting skill score was thus high?

Figure 2: The red contour lines don’t add much to the panels but make them busy and cluttered, I suggest removing the contour lines or keeping them only in one panel.

Figure 3: Panel (a) shows tracked cells, their size, and lifecycle in a very inefficient manner. I see little utility in panel (a) in its current form for the following reasons:

• Why are squares used to represent cell size when tobac can provide polygons based on the cell mask representing cell area from the segmentation step?
• Plotting all cells overlapped on top of each other presents no use. Perhaps it shows only cell 3 from the left cluster and a single cell of interest from the right cluster? Or even showing multiple panels would be better?
• Cell locations can be better represented by lines showing the cell tracks or just translucent, partly overlapping polygons colored by lifecycle stage.
• There is virtually no way to distinguish which cell is Cell 3 within that cluster despite knowing that it is meant to be the largest one.

Figure 3: Panel (b) is a very nice and useful figure. I suggest making this a standalone panel given the issues with panel (a). A minor suggestion – is there a way to distinguish between thermals tracked from near cloud base versus thermals only identified at some height above cloud base (since this is also mentioned explicitly in the text and seems an important distinction).

• There are many instances of overlapping lines in this panel. Does that mean that there were multiple thermals at the same altitude at the same time, but at a different horizontal location?
• If the authors can think of a good way to represent these thermals visually in a 3D field, that would be very useful. Alternatively, a statistical time series of the number of thermals and their relative distance could be added.
• It seems there is some correlation between the mass flux values and the base of the corresponding thermals. Darker red lines seem to start at higher altitudes. Do the authors have any comment on this? The high mass flux values likely represent high upward vertical velocity values or are they related to the thermal sizes? Could it be a tracking issue/artefact?

Line 261: Mention section where the discussion resumes for the vertical velocity differences.

Line 274: Do you mean “polluted” case instead of “clean”?

Figure 6: I admit there are already a lot of panels here but adding a panel on temperature or markers for the freezing level in one of the mixing ratio panels will provide further context on the hydrometeor phase.

Line 341: Given that the thermals were tracked in both space and time using tobac, and their lifecycles can be estimated based on the tracking times, is it not possible to create vertical composites such as those in Figure 6 where the difference between the clean and polluted cases is plotted for all thermal properties as a function of the normalized thermal lifetimes? The shading or thickness of lines could change as the lifetime increased, and that would demonstrate the lifecycle evolution of the thermal properties.

Line 362: Again, since thermals and the convective cores are tracked and their spatial coverage easily identified, I am wondering why full-domain-means are used in Figure 9, would it not be more useful to set some distance threshold beyond the cloud boundaries to consider the environmental conditions? This is especially important for the hypothesis regarding the impact of droplet/rain evaporation rates on dQv and dT.

Line 370: I’m surprised to see little discussion of the positive trends in dQr, dQis, and dQgh in the last few hours of each date’s simulations. The reinitialization experiments later address how changing CCN/aerosol concentration impacts this trend but does not comment on why its observed in the first place.

Technical corrections/suggestions:

Line 26: “enhanced aerosol concentrations”?

Line 283: maintain consistency in terminology? Initiation height in text versus starting height in figure caption.

Line 376: “fate” -> “rate”?